Kanban StudiosKanban Studios
SELECTED WORK · 05

SooqRoot - the harvest.

An Arabic-first, AI-assisted farm-to-market coordination platform built for the Ministry of Climate Change and Environment. Buyers place advance produce demand, an AI translator turns it into farmer-ready instructions, and a rule-based engine pools nearby farms and allocates orders across them - all confirmed before harvest. One order. Many farms. Confirmed before harvest.

SooqRoot landing - One order. Many farms. Confirmed before harvest.
AR ⇄ ENArabic-first UI that flips the whole layout RTL ↔ LTR
Pre-harvestOrders confirmed and allocated before a single crop is picked
Many farmsOne buyer order pooled and split across nearby producers
THE BUYER

Free text in, a confirmed spec out.

A buyer pastes a messy request. The AI translator reads it, returns a structured demand table with interpretation and confidence, and tracks every order from request to delivered - alongside live local-sourcing, shortfall, and waste-risk metrics.

SooqRoot buyer dashboard - AI-interpreted demand with structured order table
THE FARMER

Sell smarter, not alone.

Farmers declare what they can supply, see their farm profile and open harvest windows, and ask the Arabic copilot for advice - pooling, grade, packaging, and timing - the same screen, captured here in both light and dark.

SooqRoot farmer dashboard - light mode
SooqRoot farmer dashboard - dark mode
THE OPERATOR

One order, allocated across many farms.

The operator console runs the allocation engine on the demand pool - splitting each order across farms, flagging shortfalls, proposing backups and substitutes, and scoring fulfillment risk. Shown here fully in Arabic, right-to-left.

SooqRoot operator dashboard - the allocation engine in Arabic, right-to-left
WHAT IT DOES

Demand, translated. Supply, pooled.

From a buyer's plain-language request to a confirmed, multi-farm harvest plan - an AI translator, a deterministic allocation engine, a risk model, and an Arabic-first copilot, working end to end.

01

AI demand translator

Buyers type what they need in plain language. A rule-based natural-language parser turns it into a structured demand table - crop, grade, quantity, packaging, delivery window - and returns an interpretation, a confidence score, and the reasoning behind every line.

02

Rule-based allocation engine

A deterministic solver matches each order to farms by grade fit, supply confidence, location, and distance - then splits a single order across Al Akhdar, Desert Leaf, and Oasis Fresh when no farm can cover the volume alone, surfacing exact shortfalls as it goes.

03

AI fulfillment risk scoring

Every allocation is scored for fulfillment risk with written reasons and concrete mitigations, backed by automatic backup-farm fallbacks and a substitute advisor that proposes alternative produce when supply runs thin.

04

Arabic-first farmer copilot

An Arabic-native copilot advises farmers on whether to join a pool, which grade and packaging to target, and when to harvest - referencing live open demand so the guidance maps to what the market is actually asking for.

05

Harvest instruction cards

Once an order is allocated, each farm receives harvest-instruction cards plus a quality-and-packaging guide tailored to vegetables, fish, or honey - turning an abstract market signal into farmer-ready actions.

06

Batch passport traceability

Each batch carries a QR-style passport with end-to-end traceability and a local-sourcing tracker, so every delivery can be traced from a specific farm back to the original buyer demand.

BUILT WITH
REACTTYPESCRIPTTAILWIND CSSRULE-BASED AIRECHARTSARABIC-FIRST RTLVERCEL
Chat on WhatsAppWhatsApp